20 determining human-centered parameters of ergonomic micro-gesture interaction for drivers using the theater approach

1、Studying Car and DriverDetermining Human-Centered Parametersof Ergonomic Micro-Gesture Interaction for Drivers Using the Theater ApproachAngela MahrAutomotive Group, German Research Center for Artificial Intelligence (DFKI)Saarbrcken, Germanyangela.mahrdfki.deChristoph EndresAutomotive Group, German

2、 Research Center for Artificial Intelligence (DFKI)Saarbrcken, Germanychristoph.endresdfki.deTanja SchneebergerAutomotive Group, German Research Center for Artificial Intelligence (DFKI)Saarbrcken, Germanytanja.schneebergerdfki.deChristian Mller Automotive Group, German Research Center for Artificia

3、lIntelligence (DFKI) & Action Line Intelligent Transportation Systems, EIT ICT labs, Germanychristian.muellerdfki.deABSTRACT1. INTRODUCTIONDuring the last decades, the awareness regarding safe driver interfaces has constantly been increasing. With the rising number of functions integrated, tradition

4、al knob-based interaction has been enriched with more and more complex displays (sometimes touch screens), with speech interaction, and with tactile feedback. The new technologies have been developed and integrated, in order to use the drivers cognitive capabilities as efficiently as possible. Aimin

5、g at further improvement for the driver interacting with the car, several car manufactures are carrying out research on gesture recognition in collaboration with universities and research titutes 2. Gestures represent a comfortable addition to existing interaction types, without the decline in recog

6、nition accuracy under noisy driving conditions that speech still suffers from 3. Recently, several technologies for touch-free gesture recognition have been identified and gradually been improved: With vision-based systems, one or multiple cameras capture parts of the drivers body, mostly in the cen

7、ter-stack area. There arealso a number of systems based on ultrasonic sound 4.Our approach, called “Geremin”, belongs to the category of capacitive systems detecting the presence of a human hand near a conductive object. In a number of studies, this technique has been proven to be unaffected by ligh

8、t and dynamic backgrounds (like the vision-based approach) while having fast response times 5. The name is derived from the words “Gesture” and “Theremin”In this paper, we describe a techniqueto determine userpreferences concerning in-car micro-gesture interaction. Theapproach is derived from the th

9、eater technique 1, and implies a collaborative adjustment of parameters with the experimenter, until the subject has decided about the final settings. We evaluated three systematically selected gestures (zooming, sweeping, and circling) for controlling four exemplary comfort functions of the car (wi

10、ndow lifter, air condition, radio volume, and seat heating). The main result of our study is the geometry of a “sweet spot” for micro-gesture recognition close to the steering wheel, which is independent from the underlying technical recognition approach. Additionally, preferred sizes, angles, and p

11、ause times for the investigated gestures are provided. We give an indication, which of the gestures is preferred by the users (the sweeping gesture). Finally, we provide a more detailed view on the interaction between gesture preferences and function.Categories and Subject DescriptorsH.5.2 Informati

12、on Interfaces and Presentation: User Interfaces Theory and methods, Input devices and strategiesH1.2 Models and Principles: User/Machine Systems Human factorsGeneral TermsAlgorithms, Measurement,Performance, Design, Reliability, Experimentation, Human Factors(an early capacitive musical inventor Lon

13、 Theremin).trument named after the RussianThe study presented here, adds another piece to the “Geremin puzzle” by addressing the issues of the human-centeredKeywordsMicro-Gestures, Contact-Free, Modality, Parameters, Usability, Ergonomics, Theater Approach, Steering Wheel, Driving.parameters. Please

14、 note, however, that the deliveredights onsizes, angles, and preferred gesture type are likewise useful forother technical micro-gesture recognition approaches. The only prerequisite that those other approaches need to share with ours is targeting “micro-gestures” small, one-handed gestures by the d

15、river, controlling the cars comfort functions (or doing otherPermission to make digital or hard copies of all or part of this work for personal or classroom use is granted without fee provided that copies are not made or distributed for profit or commercial advantage and that copies bear this notice

16、 and the full citation on the first page. To copy otherwise, or republish, to post on servers or to redistribute to lists, requires prior specific permission and/or a fee.AutomotiveUI 2011, Nov. 30th Dec. 2nd 2011, Salzburg, Austria Copyright 2011 ACM 978-1-4503-1231-8/11/11. $15.00151AUI2011 Procee

17、dingsSalzburg, Austria, Nov. 30th Dec. 2nd 2011tertiary tasks), without the need to take the hand off the steering wheel.2. THE THEATER APPROACHWith a user-centric design approach, it is beneficial to let the user actively participate early in the design process. For example, in a Wizard of Oz (WoOz

18、) experiment, subjects interact with a computer system that they believe to be autonomous, but which is actually being operated or partially operated by an unseen experimenter (the “wizard”). A subject may think to be communicating with a computer using a speech interface, when the wizard in another

19、 room is secretly translating the participants words into regular computer commands. The Theater-system technique extends the WoOz technique in a way that the experimenter, the so-called confederate, is visible to the subject. Subject and confederate play through different use cases, as if they woul

20、d play a role in a theater. While the WoOz technique is used for the evaluation of functionality, the theater approach can be used both for evaluation and design. The theater- system technique was proposed by 1 for designing a haptic- multimodal interaction strategy for highly automated vehicles. We

21、 adopt the technique here, for determining the human-centered parameters for our proposed micro-gesture interface.3. TECHNICAL IMPLEMENTATION OF MICROGESTURE RECOGNITIONThe objective of the experiment described in this paper, was to specify parameters necessary for building an in-car micro-gesture r

22、ecognition device. In order to provide a better understanding of the context, we shortly introduce our Geremin device here 5. The empiric results presented, however, are generic in nature and are not depending on any specific, underlying recognition technology.The Geremin is a recognizer for micro-g

23、estures in an in-car environment particularly for the driver. Its aim is to provide an additional modality to control (comfort) functions of the car, without the need for taking the eyes off the road or the hands off the wheel. It is based on electric field sensing and attached in close proximity to

24、 the right (and/or left) hand of the driver at the 2 oclock position (respectively “10 oclock” position) of the11.6% of the VR users used interaction with hand gestures. Based on that study, we claim that gestures as input modality are a suitable extension of the standard in-car interaction. This is

25、 also supported by 6, who state, a gesture interface is a viable alternative for completing secondary tasks in the car. According to 2, gestures are originally not self-revealing and therefore need explanation and visual reminders. At the same time, the authors acknowledge that providing visual remi

26、nders would necessarily neutralize any potential safety benefit. Nevertheless, 7 expects numerous automotive applications by 2020.According to 2, three different application doma automotive hand gestures have been addressed:for mapping1. Direct mapping of gestures to the complete functionality of in

27、- vehicle devices (e.g. radio, CD, navigation system). Although this approach could lead to a very consistent interface, the authors conclude that too many gestures would be needed and thus many of them would not be natural.2. Mapping to in-vehicle controls, mimicking each individual control type (p

28、ush button switch, push and hold button switch, rotary position selector, etc.). This type is not recommended as well by 2, because creating natural mimic gestures for each control type has substantial limitations like transferability between different cars.3. According to the authors, the third cat

29、egory, selective mapping to theme or function, appears to have the most realistic practical possibilities. With this approach, a relatively small selection of gestures is mapped consistently to a (again relatively small) number of functions. 8 provides examples belonging to this category: a) waving-

30、off incoming calls; b) using ones index finger with a clockwise/counterclockwise rotation to raise respectively lower the stereo volume. 9 proposes skipping between music titles, albums, radio stations or enabling/disabling audio sources.As indicated above, we envision testing the selective mapping

31、of a limited set of micro-gestures performed in the immediate area of the steering wheel without taking a hand off. The question, which different kinds of micro-gestures can still be performed in an ergonomic way while holding the steering wheel, has recently been investigated 10. Independent of any

32、 technical solutions, the authors present 17 gestures that can be performed with varying effort, while the palm is grasping something with moderate pressure. This was a rich basis for the selection of micro-gestures for our study.5. PREDEFINITION OF GESTURE TYPESIn the beginning, the micro-gestures

33、and tasks for our study had to be chosen from a variety of possibilities. To start with the selection of the micro-gestures, there were a number of aspects to considered. On the one hand we considered the ergonomic aspects of the context that the driver should still be able to hold the steering whee

34、l at least with the palm and the remaining fingers. On the other hand gesture selection is limited by the ease of performance: The driver would perform the gesture from time to time and should not feel uncomfortable or even painful applying the gesture. Some gestures one could think of would also be

35、 too hard to apply like for example moving the ring and index fingers simultaneously. Without any training a gesture like this is hard to perform for most people 10. Drivers should furthermore be able to a) learn the gesture easily and b) recall the learned gesture when needed. Also the risk of gest

36、ures being mixed up with natural or accidental movements should be kept to a minimum.steering wheel. Geremancient ancestor is an early electronicmusictrument, invented and patented by Lon Theremin in the1920ies. Like the “Theremin”, our system uses the combination of antenna and hand as a capacitor,

37、 and detects changes in the distance of the hand. However, it does not oscillate and generate sound, but directly converts changes in voltage into numbers to be transmitted via a USB interface. The resulting data is thereupon processed on the computer using state-of-the-art machine learning algorith

38、ms, e.g. Dynamic Time Warping (DTW), or most recently with custom developed geometric analysis methods. Furthermore, the Geremin is capable of using several antennae, i.e. three in the most recent prototype.4. RELATED WORKIn addition to the assessment of micro-gesture settings and the proof of conce

39、pt of a new technology, drivers acceptance of the new technology needs to be considered. Would micro-gestures be an accepted additional input modality? 3 compared a traditional 2D in-car interface with an innovative 3D Virtual Reality (VR) interface in a user study in order to evaluate performance a

40、nd general acceptance. Both interfaces could be used with traditional controls (switches, buttons) as well as with gestures and speech commands. 15.5% of the users of the traditional interface and152Studying Car and DriverTo sum up, together with some technological considerations these are the reaso

41、ns why we decided to engage a maximum of two fingers for the gestures for a start. Moreover, we wanted to make use of analogies to well-known gestures like the ones used for the iPhone interaction and transfer them to this new context. In an informal bra torming session (4 researchers familiar with

42、the field) we resumed former ideas for micro-gestures, generated new ones and rated all of them. Finally we chose the most promising three gestures for our study: the zooming gesture, the sweeping gesture, and the circling gesture.The aim of the experiment presented here was to answer the following

43、questions: 1. Do users find such a system useful? 2. Which gesture is preferential to the users for +/- kind of controls: circle, sweep, or zoom? 3. What is the set of parameters for finding a mapping between the (recognized) motion of the subjects finger and the status of a device?A computer flat s

44、creen (size 19 inch, resolution 1280 x 1024 pixel) was set up behind the steering wheel in a distance of about 8 centimeters to the position where the subjects right hand was holding the wheel. The distance between the steering wheel and the screen was individually adjusted to index finger length pl

45、us about 1 centimeter. Accordingly, subjects were close to the screen, but did not touch it. On the screen our experimental gesture-program displayed a moving red dot that indicated the position of the subjects right index finger (sometimes in combination with the thumb) and hence the gesture to be

46、performed. The subject should follow the red dot as exactly as possible. A screenshot of this interface for the example of a sweeping gesture can be seen in Figure 2.6. EXPERIMENT6.1 Subjects24 subjects participated in this experiment, 12 men and 12 women. Their age varied between 21 years and 67 ye

47、ars (mean = 37.0, SD = 15.1). All of them possessed a valid drivers license for at least two years, and 86 percent of them were driving regularly. We just allowed right- or two-handed people to our study, as our setup had so far been implemented in the right- handed version only. We measured subject

48、s hand and body size, to check for any correlation with gesture preferences. Body size varied between 1.59 and 1.85 meters (mean = 1.70, SD = 0.08), gesture span measured as maximal distance from the tip of the index finger to the tip of the thumb varied between 12 and 20 centimeters (mean = 15.4, S

49、D = 2.2). We furthermore determined subjects length of the index finger only. It varied between 6 and 8 centimeters (mean 7.2, SD = 0.7). All subjects were native German speakers. They were paid 8 Euros for approximately 1 hour of time.6.2 ApparatusThe experiment setup consisted of a steering wheel

50、attached to a desk in a way that resembles the ergonomic condition of a driver in a car (cf. Figure 1). During the experiment the subject was holding the steering wheel with both hands at the 10/2 oclock position.Figure 2: Screenshot of the program, which presented the micro-gestures to be performed

51、 (the example shows the sweeping gesture). The right index finger should follow the position of the red dot moving along the line. With the help of the sliders on the right hand side the gesture parameters could be adapted.On the right upper side there were sliders with which each of the gestures co

52、uld be configured in various ways (size, location, speed, etc.). A detailed description of the settings for each gesturecan be foundection 6.4. By using a mouse the experimenteradapted the settings according to the demands of the subject. The effect of the gesture performed with the selected setting

53、s for a special task was shown on a second screen located on the subjects right hand side.6.3 Exemplary Interaction TasksAs we chose to refer to the “selective mapping” 2 (see above), our goal was to select one suitable micro-gesture for a larger number of tasks even though they are performed and ob

54、served in a completely different fashion. As a result, we wanted to choose a specific set of tasks for the experiment that is as representative as possible for a big variety of in-car comfort functions. First of all the majority of in-car (comfort) tasks comprises an increasing or decreasing of an a

55、ctual state of a device to a desired level,whereas the number of levels that can be selected is an important factor that so far usually becomes obvious in the specific design of buttons and knobs with their labels and tactile feedback. Accordingly, we selected four exemplary in-car devices to be con

56、trolled by the subjects via micro-gestures: The window lifter, the air condition, the radio volume, and the seat heating (cf. Figure 3).Figure 1: Setup153AUI2011 ProceedingsSalzburg, Austria, Nov. 30th Dec. 2nd 2011Figure 4: Demonstration of the “zoom” gesture.Figure 3: Animated controlled devices a

57、s gesture feedback to the subject.In the main experimental block subjects completed one of the four interaction tasks after the other (order was counterbalanced between subjects). For every interaction task subjects investigated each of the three gesture types (order also balanced between subjects).

58、 For each gesture the subject had the experimenter adapt the parameters (with the help of a mouse) until she was satisfied with the result. Each task was performed from the minimum to the maximum, the maximum to the minimum and also for a partial range of about one third in order to achieve results

59、that are valid for the complete range but also for fine tuning of functions. After the subject had decided about her final settings, the experimenter recorded the final parameters for this gesture and task. The program furthermore delivered the number of gesture iterations performed and the total task tim